Above: Location of Ewa Beach on O`ahu
Above: Ewa Plantation as an active mill
Ewa Plantation began in 1890 as a sugarcane plantation mill. Through the 1920s and 1930s, Ewa Plantation was a booming and growing business. Perhaps the plantation began monitoring climate because of climate’s impacts on sugarcane, but in any case, they had a climatic dataset beginning in 1905. Currently, sugar plantations in Hawai`i are no longer in operation; therefore, the Ewa Plantation now serves as a historic landmark. However, they continued collecting climatic data even after they were no longer a functioning plantation, collecting data up until the year 2006.
After downloading their dataset off of NOAA’s climate database for the Ewa Plantation station on the island of O`ahu, we inputted it into R for processing. This dataset includes variables like maxmimum temperature, minimum temperature, and precipitation. This blog will explore trends in temperature and precipitation at this specific station.
When we first import the data, it is not yet in a form that we can use and visualize well. As you can see, there are several values that don’t make sense. How could the maximum temperatures be almost -10000 degrees? It’s actually not–the value -9999 is used to represent having no data. We can change these values to “NA” to ensure that the big negative values do not skew our data.
After changing the -9999 values to NA, we can check to see if our data is now usable:
Even though we removed the missing values, the data is still unusable. The reason why has to do with how data used to be recorded. Instead of chronologically entering data in a way that Excel could read and turn into a continous timeline, dates used to be inputted in the format of YYYYMMDD. This means that, for example, 19131231 (December 31, 1913) would then turn into 19140101 (January 1, 1914). Chronologically, there are several missing values in between which is why there are so many gaps. The data still needs to be chaned into a form in which dates are continuous and numerically consecutive. In this way, it is true scientists have to manipulate data–but it is not so much the content of the data as the format.
To get the data to organize like we want, we first need to convert the date to string or character values. After that, we can convert the strings to a data format.
## [1] "19050101" "19050102" "19050103" "19050104" "19050105" "19050106"
## Subset Sites
I have two sites in this data, so, we need to subset it.
## [1] EWA PLANTATION 741 HI US
## Levels: EWA PLANTATION 741 HI US
Let’s choose EWA PLANTATION 741 HI US as the station name…
##
## Call:
## lm(formula = TMAX ~ NewDate, data = Ewa)
##
## Residuals:
## Min 1Q Median 3Q Max
## -19.0590 -2.5163 0.3007 2.9257 9.1843
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.414e+01 3.286e-02 2560.28 <2e-16 ***
## NewDate 3.097e-05 2.645e-06 11.71 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.589 on 25580 degrees of freedom
## (9006 observations deleted due to missingness)
## Multiple R-squared: 0.00533, Adjusted R-squared: 0.005291
## F-statistic: 137.1 on 1 and 25580 DF, p-value: < 2.2e-16
As shown, the maximum temperature annually has had a slight upward trend from 1905 when the records began until 2006. The slope is about 3.1e-05 (a very tiny slope). This means that, according to the trend, maximum temperatures increase about .000031 degrees per year, or .0031 degrees per century. This increase in maximum temperature is very statistically signficant, with a p-value of 2.2e-16 (a p-value of 0.05 or less would be statistically significant). This means that we can reject the null hypothesis that there is no relationship between change in temperature and time. However, we can see that there is a lot of missing data beginning at roughly 1980. While I tried to look into why there were no available measurements for about a 20 year span, no plausible explanations were offered online.
But since there’s a lot of missing data beginning at about 1980, let’s focus on the trend from 1905 up until that point to see if the missing data has an effect on the trend slope.
Here is the graph of maximum temperatures from 1905 up to 1978:
Now, let us analyze the trend of the maxmimum temperature data from 1905 to 1978.
##
## Call:
## lm(formula = TMAX ~ NewDate, data = Ewa[Ewa$DATE < 19780101,
## ])
##
## Residuals:
## Min 1Q Median 3Q Max
## -19.4039 -2.4252 0.3501 2.7643 9.1120
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 8.457e+01 3.892e-02 2172.87 <2e-16 ***
## NewDate 6.565e-05 3.132e-06 20.96 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 3.539 on 24399 degrees of freedom
## (1565 observations deleted due to missingness)
## Multiple R-squared: 0.01769, Adjusted R-squared: 0.01765
## F-statistic: 439.4 on 1 and 24399 DF, p-value: < 2.2e-16
The trend of maxmimum temperatures has a steeper slope of 6.565e-05 when excluding all of the missing data, indicating a .0000656 degree increase each year, or a .00656 degree increase per century. This increase is also statistically significant, with a p-value of 2.2e-16.
Let’s do the same with minimum temperature measurements.
##
## Call:
## lm(formula = TMIN ~ NewDate, data = Ewa[Ewa$DATE < 19780101,
## ])
##
## Residuals:
## Min 1Q Median 3Q Max
## -18.1094 -2.5150 0.4151 2.7712 14.6404
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 6.560e+01 4.403e-02 1489.90 <2e-16 ***
## NewDate 5.783e-05 3.517e-06 16.45 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4.012 on 24531 degrees of freedom
## (1433 observations deleted due to missingness)
## Multiple R-squared: 0.01091, Adjusted R-squared: 0.01086
## F-statistic: 270.5 on 1 and 24531 DF, p-value: < 2.2e-16
As you can see, minimum temperatures also have an upward trend, with a slope of about 5.8e-05. This translates into a .000058 degree increase per year, or a .0058 degree increase each century. This increase in minimum temperatures is also statistically significant, with a p-value of again 2.2e-16.
While Hawaii’s temperatures are relatively constant year-round, let’s still take a look at individual months and their data across different years.
Let’s now look at the trend of temperatures in January throughout the years.
##
## Call:
## lm(formula = TMAX ~ YEAR, data = MonthlyMean[MonthlyMean$Month ==
## "01", ])
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.7484 -1.0961 0.1105 1.1762 3.5595
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 46.885795 15.929432 2.943 0.00432 **
## YEAR 0.016999 0.008184 2.077 0.04123 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.839 on 75 degrees of freedom
## Multiple R-squared: 0.05439, Adjusted R-squared: 0.04178
## F-statistic: 4.314 on 1 and 75 DF, p-value: 0.04123
Now, the change is 0.0169988 degress/year or 1.7 degress/100 years with a p-value of 0.0412.
Now let’s try the same thing with November.
##
## Call:
## lm(formula = TMAX ~ YEAR, data = MonthlyMean[MonthlyMean$Month ==
## "11", ])
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.4423 -0.6529 -0.0859 0.8901 2.6930
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 57.277543 12.896501 4.441 3.12e-05 ***
## YEAR 0.013417 0.006629 2.024 0.0466 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.402 on 73 degrees of freedom
## Multiple R-squared: 0.05313, Adjusted R-squared: 0.04016
## F-statistic: 4.096 on 1 and 73 DF, p-value: 0.04664
Now, the change is 0.0134173 degress/year or 1.342 degress/100 years with a p-value of 0.0466.
First we should create monthly mean for each month.
## Month Year TMIN YEAR
## 1 01 1905 55.88000 1905
## 2 02 1905 57.33333 1905
## 3 03 1905 58.55556 1905
## 4 04 1905 62.64000 1905
## 5 05 1905 65.66667 1905
## 6 06 1905 66.32000 1905
Here’s a graphic showing the trend of minimum temperatures in November over the years:
##
## Call:
## lm(formula = TMIN ~ YEAR, data = MonthlyMeanTMIN[MonthlyMeanTMIN$Month ==
## "11", ])
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.7474 -1.1263 0.2163 1.3812 2.7583
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 21.326505 15.299028 1.394 0.16750
## YEAR 0.022462 0.007866 2.855 0.00557 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.687 on 74 degrees of freedom
## Multiple R-squared: 0.09925, Adjusted R-squared: 0.08708
## F-statistic: 8.154 on 1 and 74 DF, p-value: 0.005573
Let’s look at the trends of minimum temperature throughout the years for the month of July.
##
## Call:
## lm(formula = TMIN ~ YEAR, data = MonthlyMeanTMIN[MonthlyMeanTMIN$Month ==
## "07", ])
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.93102 -1.08270 0.06168 1.07673 3.13611
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 20.397887 12.967313 1.573 0.119923
## YEAR 0.024659 0.006665 3.700 0.000409 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.485 on 75 degrees of freedom
## Multiple R-squared: 0.1544, Adjusted R-squared: 0.1431
## F-statistic: 13.69 on 1 and 75 DF, p-value: 0.0004088
The minimum temperatures in the month of July are increasing as well, with a slope of .025. This means that at the current trend, temperatures are increasing at a rate of .025 degrees fahrenheit per year, or 2.5 degrees fahrenheit per century. The understood goal is to keep climate change within 2 degrees celcius or 3.6 degrees fahrenheit above preindustrial levels, and while we can by no means generalize the trends of Ewa Station to the entire globe, this is still a concerning rate.
Here we will try to analyze trends in rainfall at Ewa Plantation throughout the years.
## Month Year PRCP YEAR
## 1 01 1905 0.006451613 1905
## 2 02 1905 0.000000000 1905
## 3 03 1905 0.018064516 1905
## 4 04 1905 0.007000000 1905
## 5 05 1905 0.005161290 1905
## 6 06 1905 0.110000000 1905
Let’s try looking at trends in precipitation during specific months. Let’s take a look at the month of May.
##
## Call:
## lm(formula = PRCP ~ YEAR, data = MonthlyMeanPRCP[MonthlyMeanPRCP$Month ==
## "05", ])
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.07182 -0.03716 -0.02345 0.00010 0.47633
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.3268074 0.6141454 2.160 0.0333 *
## YEAR -0.0006561 0.0003144 -2.087 0.0397 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.08609 on 92 degrees of freedom
## Multiple R-squared: 0.0452, Adjusted R-squared: 0.03483
## F-statistic: 4.356 on 1 and 92 DF, p-value: 0.03965
While the data at the Ewa Plantation station is incomplete, it does undoubtedly point to a trend that we would expect to come with climate change. Both the minimum and maximum temperature trends increase over the years. Also, both months in the summer as well as in the winter see this upward climbing trend.
Though the trend line is not very steep, any change has to be considered over long periods of time. If the temperature follow along this same trend for a much longer period of time, the temperature could then have an overall significant increase. Also, this data unfortunately does not include data from the most recent years, so it could be missing trends in temperature caused by increases in carbon dioxide levels since 2006. If the station had included readings from 2015 and onward, we could possibly see the effects of carbon dioxide in the atmosphere reaching over 400 parts per million for the first time in recorded history.